Skip to main content

Tech Stack

What the skAInet Edge-Compute is designed, built, and run with. We prefer open tools and open formats wherever they exist β€” so that what runs on your Edge-Compute is inspectable, and what you build on it is portable.

Hardware design & manufacturing​

Tool / partnerUsed for
🧩KiCadAll schematic and PCB designs β€” carrier board, PoE switch board, and every measurement device
πŸ“build123dParametric CAD for the anodized aluminum enclosure, written as Python code
🏭HLH PrototypesManufacturing partner for the machined and anodized enclosures
πŸ”ŒGH ElectronicsCertified M12 PoE cable assemblies (Cat-5e, M12 X-/D-coded)
🟒NextPCBPCB fabrication and assembly (PCBA)

On the Edge-Compute​

ToolUsed for
🐧Yocto ProjectThe Linux distribution β€” a reproducible image with a documented SBOM, aligned with the EU Cyber Resilience Act
πŸ”NetBirdSecure, WireGuard-based remote access for maintenance and support (optional)
🚏RedisData transport on the device β€” measurement streams flow through Redis between mappers, aggregators, and APIs
🐳DockerDeployment unit for on-edge analytics; your analysis ships as a container
⚑CrowHigh-performance C++ web framework for the device's REST and WebSocket APIs
πŸ…°οΈAngularFrontends served directly from the Edge-Compute

External integrations (optional)​

The Edge-Compute works standalone. When connected to upstream infrastructure, it integrates with:

ToolUsed for
πŸ—„οΈCephPermanent data storage β€” aggregated data is written to an external Ceph cluster
☸️KubernetesHigh-performance analytics that exceed the device's own compute budget
🧠OllamaLocal large-language-model inference for data insights β€” on your infrastructure, not a cloud

Protocols & fieldbus​

ProtocolUsed for
πŸ”¬EPICSIntegration into experimental-physics control systems (e.g. the PANDA luminosity detector at GSI/FAIR)
πŸ—οΈModbus TCPIntegration into plant networks, PLCs, and SCADA systems

For the full list of upstream API options β€” REST, WebSocket, Webhooks, MQTT, OPC UA, gRPC, and more β€” see the platform overview.

Why this stack​

  • Reproducible: Yocto builds the same image every time, and the SBOM lists every package in it.
  • Portable: Your code runs in Docker containers against a Redis stream β€” it does not care which compute module is inside the enclosure.
  • Open: KiCad, build123d, Yocto, Redis, Docker, Crow, Angular, Ceph, Kubernetes, Ollama, EPICS, and Modbus are all open source or open standards.
  • Serviceable: With NetBird enabled, Auto-Intern can support and update your device remotely β€” and you can switch it off at any time.